DNA computing is a new method that uses the DNA molecule as a computing medium and its biochemical reactions as computing tools. In DNA computation, the core reaction is specific hybridization, and effective encoding can reduce false hybridization. In this paper, DNA thermodynamic functions are transformed into proper constrained terms, and the problem of DNA encoding is then regarded as a multi-objective optimization problem. Combined with three conventional constrained terms, this multi-objective problem is solved by multi-objective particle swarm optimization, and better DNA sequences are found. By comparing with the sequences of a single-objective particle swarm and conventional sequences, the validity and advantage of our method are proven.